From ca5e10e2dc56e2ba45a4a1ebab003438d458dafd Mon Sep 17 00:00:00 2001
From: barneygale <barney.gale@gmail.com>
Date: Fri, 26 May 2023 21:27:23 +0100
Subject: [PATCH 1/3] GH-104996: Defer joining of `pathlib.PurePath()`
 arguments.

Joining of arguments is moved to `_load_parts`, which is called when a
normalized path is needed.
---
 Lib/pathlib.py                                | 33 +++++++++++--------
 ...-05-26-21-24-06.gh-issue-104996.aaW78g.rst |  2 ++
 2 files changed, 22 insertions(+), 13 deletions(-)
 create mode 100644 Misc/NEWS.d/next/Library/2023-05-26-21-24-06.gh-issue-104996.aaW78g.rst

diff --git a/Lib/pathlib.py b/Lib/pathlib.py
index 8cb5279d735a30..59907b56e27d1b 100644
--- a/Lib/pathlib.py
+++ b/Lib/pathlib.py
@@ -244,9 +244,9 @@ class PurePath(os.PathLike):
     """
 
     __slots__ = (
-        # The `_raw_path` slot stores an unnormalized string path. This is set
+        # The `_raw_paths` slot stores unnormalized string paths. This is set
         # in the `__init__()` method.
-        '_raw_path',
+        '_raw_paths',
 
         # The `_drv`, `_root` and `_tail_cached` slots store parsed and
         # normalized parts of the path. They are set when any of the `drive`,
@@ -299,10 +299,11 @@ def __init__(self, *args):
         paths = []
         for arg in args:
             if isinstance(arg, PurePath):
-                path = arg._raw_path
                 if arg._flavour is ntpath and self._flavour is posixpath:
                     # GH-103631: Convert separators for backwards compatibility.
-                    path = path.replace('\\', '/')
+                    paths.extend(path.replace('\\', '/') for path in arg._raw_paths)
+                else:
+                    paths.extend(arg._raw_paths)
             else:
                 try:
                     path = os.fspath(arg)
@@ -313,13 +314,8 @@ def __init__(self, *args):
                         "argument should be a str or an os.PathLike "
                         "object where __fspath__ returns a str, "
                         f"not {type(path).__name__!r}")
-            paths.append(path)
-        if len(paths) == 0:
-            self._raw_path = ''
-        elif len(paths) == 1:
-            self._raw_path = paths[0]
-        else:
-            self._raw_path = self._flavour.join(*paths)
+                paths.append(path)
+        self._raw_paths = paths
 
     def with_segments(self, *pathsegments):
         """Construct a new path object from any number of path-like objects.
@@ -349,7 +345,14 @@ def _parse_path(cls, path):
         return drv, root, parsed
 
     def _load_parts(self):
-        drv, root, tail = self._parse_path(self._raw_path)
+        paths = self._raw_paths
+        if len(paths) == 0:
+            path = ''
+        elif len(paths) == 1:
+            path = paths[0]
+        else:
+            path = self._flavour.join(*paths)
+        drv, root, tail = self._parse_path(path)
         self._drv = drv
         self._root = root
         self._tail_cached = tail
@@ -673,7 +676,11 @@ def is_absolute(self):
         # ntpath.isabs() is defective - see GH-44626 .
         if self._flavour is ntpath:
             return bool(self.drive and self.root)
-        return self._flavour.isabs(self._raw_path)
+        else:
+            for path in self._raw_paths:
+                if self._flavour.isabs(path):
+                    return True
+            return False
 
     def is_reserved(self):
         """Return True if the path contains one of the special names reserved
diff --git a/Misc/NEWS.d/next/Library/2023-05-26-21-24-06.gh-issue-104996.aaW78g.rst b/Misc/NEWS.d/next/Library/2023-05-26-21-24-06.gh-issue-104996.aaW78g.rst
new file mode 100644
index 00000000000000..8b81b681af94aa
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2023-05-26-21-24-06.gh-issue-104996.aaW78g.rst
@@ -0,0 +1,2 @@
+Improve performance of :class:`pathlib.PurePath` initialisation by
+deferring joining of paths when multiple arguments are given.

From d528a0aecfa2f2b4b96eeb4353bbab1e7092ec9e Mon Sep 17 00:00:00 2001
From: barneygale <barney.gale@gmail.com>
Date: Wed, 7 Jun 2023 22:00:21 +0100
Subject: [PATCH 2/3] Tighten up is_absolute()

---
 Lib/pathlib.py | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/Lib/pathlib.py b/Lib/pathlib.py
index 59907b56e27d1b..87f904e02a0132 100644
--- a/Lib/pathlib.py
+++ b/Lib/pathlib.py
@@ -673,14 +673,17 @@ def parents(self):
     def is_absolute(self):
         """True if the path is absolute (has both a root and, if applicable,
         a drive)."""
-        # ntpath.isabs() is defective - see GH-44626 .
         if self._flavour is ntpath:
+            # ntpath.isabs() is defective - see GH-44626.
             return bool(self.drive and self.root)
-        else:
+        elif self._flavour is posixpath:
+            # Optimization: work with raw paths on POSIX.
             for path in self._raw_paths:
-                if self._flavour.isabs(path):
+                if path.startswith('/'):
                     return True
             return False
+        else:
+            return self._flavour.isabs(str(self))
 
     def is_reserved(self):
         """Return True if the path contains one of the special names reserved

From 08c69bf0a123dadfb455c32216ad6ec39a8c268c Mon Sep 17 00:00:00 2001
From: barneygale <barney.gale@gmail.com>
Date: Wed, 7 Jun 2023 22:40:07 +0100
Subject: [PATCH 3/3] De-duplicate on string path.

---
 Lib/pathlib.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/Lib/pathlib.py b/Lib/pathlib.py
index 98eab482ae2c67..d8c597f1027f30 100644
--- a/Lib/pathlib.py
+++ b/Lib/pathlib.py
@@ -195,9 +195,10 @@ def _select_unique(paths):
     yielded = set()
     try:
         for path in paths:
-            if path not in yielded:
+            path_str = str(path)
+            if path_str not in yielded:
                 yield path
-                yielded.add(path)
+                yielded.add(path_str)
     finally:
         yielded.clear()